next up previous
Next: 7.2 Advection Up: 7. Model physics Previous: 7. Model physics


7.1 Timestepping

In older TM5 versions, the operator splitting sequence was implemented as:

\begin{displaymath}
(x)^n~(y)^m~(z)^l~vsc~csv~(z)^l~(y)^m~(x)^n~
\end{displaymath} (7.1)

with n, m, and l the number of iterations needed to fullfill the CFL criterium. Disadvantage was that for divergent flows, many iterations were needed or no solution was found at all (e.g. at the poles).

A new procedure was introduced in august-2005. The new procedure involves an iteration over the whole operator sequence, thus including vertical mixing, chemistry, and sources/sinks :

\begin{displaymath}
(xyz~vsc~csv~zyx)^l
\end{displaymath} (7.2)

The number of iterations l is determined after reading the advective mass fluxes by the CFL criterium. The .rc file should contain the variable ndyn_max. This global timestep is reduced if a CFL violation is enxountered, After reading new advective fluxes the global timestep is reset to ndyn_max.

Consequnces: Evaluation of mean quantities should consider the fact that the time step is not constant. Relevant for station output, mmix output and mix output.


next up previous
Next: 7.2 Advection Up: 7. Model physics Previous: 7. Model physics
Created at IMAU on Tue Mar 3 05:00:05 CET 2009 by a slave chained in the basement